At Avvista.ai / Cato, I helped build the AI and data infrastructure for an AI-powered public procurement platform, contributing to its successful acquisition by Cato AI. I engineered procurement data ingestion across 50+ portals and a scraping agent that raised document recovery from a 0% baseline to over 90% across 34 Italian public-tender portals.
At ZOLO, I reduced Azure OCR costs by up to 70% and daily LLM inference costs by 40% through workflow and model optimizations. I also improved automated reconciliation accuracy from 70% to approximately 88% and productionized CRF models for extracting structured data from customer orders.
At Earlybird AI, I worked on accounting automation, including an autonomous reconciliation system that reached about 80% accuracy and an email-to-ledger pipeline that reduced manual processing time by about 90%. At MNC Group, I developed recommendation systems and semantic search and conversational AI for media platforms.
Earlier, I built RAG and conversational commerce systems at Jatis Mobile and 7-Network Pte Ltd, and information retrieval and document-processing systems at Kalbe Farma. At Pipeline, I built data pipelines to BigQuery and developed ML models; at United Tractors, I built an attendance app and company profile website.

